WebAIM - Web Accessibility In Mind

WebAIM Million – 2021 Update

The 2021 WebAIM Million Report is now available. This annual accessibility analysis of the home pages of the top one million web sites provides insight into the current state of and trends for web accessibility. The report provides details into technical aspects of accessibility and comparisons for many types of home page categories, such as by TLD, document language, site sector, and technologies in use. The WAVE web accessibility testing tool was used to analyze the 1,000,000 home pages.

Here are some interesting and noteworthy items from this year’s analysis:

  • The number of detectable accessibility errors was 51.4 on average per home page. This was an improvement from 60.9 errors just one year ago.
  • The number of elements on the average home page was 887, a slight increase from 864 in 2020. At 51.4 errors per page, users with disabilities would expect to encounter detectable errors on 1 in every 17 home page elements.
  • 97.4% of home pages had detectable WCAG 2 failures. This was a small improvement from 98.1% in 2020.
  • Home pages most commonly had low contrast text, missing alternative text, missing form input labels, empty links, missing document language, and empty buttons.
  • 86.4% of home pages had low contrast text averaging 31 instances per home page.
  • 26% of images had missing alternative text. Over one third of all images analyzed had detectable accessibility issues.
  • Nearly half of all form inputs were not properly labeled.
  • The proper use of headings in home pages is increasing over time.
  • ARIA usage increased 25% since 2020 with 68% of home pages utilizing ARIA and an average of 48 ARIA attributes per page. Home pages with ARIA present averaged 41% more detectable errors (24 additional potential barriers per page) than those without ARIA. An increase in ARIA attributes aligned with an increase in detectable web accessibility errors.
  • 79% of home pages had a valid HTML5 doctype. HTML5 pages had nearly double the page elements and 35% more accessibility errors than pages with other doctypes.
  • Web site categories that were subject to increased civil rights complaints and lawsuits in 2020 were among the most improved.
  • There were significant differences in detectable accessibility errors based on top-level domain. For example, .ru (Russia) and .cn (China) home pages had around double the errors as .us (United States) and .ca (Canada) home pages.
  • With the exception of React and Ruby on Rails, the presence of all other popular JavaScript frameworks, libraries, and web frameworks corresponded with an increase in detectable errors.

There is much more information and analysis available in the 2021 WebAIM Million Report.

Comments

  1. Armony ALTINIER

    Thank you so much for this great study. These data are so important to have a clear and objective view of the Web accessibility global level.

    Look forward to going deeper in studying all these data!

  2. Joseph
  3. Jorge Fernandes

    Very interesting analysis. Thanks a lot for that. I really appreciate the concept of complexity of pages (homepages). I suggest to introduce in this concept of complexity also the number of links (mainly the internal links) in the homepage.

    In the field of images, WAVE already detect alternative text in attributes as aria-label, aria-labelledby, title – as proposed ACT-Rules – or only in alt of img? And about SVG files? I note that developers are increasingly using the SVG files inline in html and not element. Do you have data about the use of svg element?

    Great Job!

  4. Jared Smith

    What do you say on this critic on your study?

    It’s a bit difficult to provide much commentary because it is based on fundamental misunderstanding of our research and accessibility at large. For example, it begins with “98 percent of most used websites not accessible.” Our research makes no assertions of “accessibility” or “inaccessibility”, but instead only provides data on detectable errors and conformance.

    “Private organizations usually aim for A, not AA.” This is not at all my experience. AA is the basis for nearly all legal requirements.

    “Whether Wave displays errors or not is completely irrelevant. A lazy but smart developer will run the tool over, iron out the errors and get his site compliant without having improved accessibility one bit.” The author suggests that WCAG conformance has no impact on actual accessibility. This is patently false. WAVE errors almost always have negative end user impact, so addressing these errors will almost always improve the end user experience.

    “A large proportion of errors are likely to be due to such embedded content: This is, for example, social media content or advertising. If you go by WebAIM, you should probably leave out such content, since you can’t make it accessible.” Content embedded within a web page impacts the user experience and conformance of that page. It is a disturbing suggestion that it should be ignored in our analysis “since you can’t make it accessible”. Clearly such content can be made accessible or could be omitted from the page if it is not.

    I could go on, but it’s a bit difficult to follow the author’s logic any further.

  5. Jared Smith

    In the field of images, WAVE already detect alternative text in attributes as aria-label, aria-labelledby, title — as proposed ACT-Rules — or only in alt of img?

    WAVE considers ARIA labeling in its alternative text analysis. If ARIA or title is present and adequately defines an accessible name for the image, this is not defined as an alternative text failure (even though alt is much preferred).

    And about SVG files? I note that developers are increasingly using the SVG files inline in html and not element. Do you have data about the use of svg element?

    We do not have these data. We do consider svg elements with role="img" to be images and test them as such to ensure they have an accessible name.

  6. Sandrine

    Hi Jared,

    Thanks very much for the survey, the article and your responses.

    I have been looking for an archive of the previous years of the WebAIM Million to see the evolution of the raw data, like for the other surveys, but can’t find them anywhere.

    Could you make them available online ?

  7. d-bass

    What does this mean?
    1. Pages with ARIA present averaged 41% more detectable errors (24 additional potential barriers per page) than those without ARIA?
    2. An increase in ARIA attributes aligned with an increase in detectable web accessibility errors?

  8. Jared Smith

    Sandrine –

    You can access the previous versions of the reports for 2019 and 2020 at https://webaim.org/projects/million/2019 and https://webaim.org/projects/million/2020

  9. Jared Smith

    d-bass –

    Pages that had at least one ARIA attribute present had 24 more errors on average than pages that did not have any ARIA attributes present.

    As the number of ARIA attributes increased on the home page, the number of detectable accessibility errors also tended to increase.